Autonomous mobile robot path planning in unknown dynamic environments using neural dynamics

Abstract

In this paper, a novel variant of bio-inspired planning algorithms is presented for robot collision-free path planning in dynamic environments without prior information. The first contribution of this paper is that, with mild technical analysis, the traditional neural dynamic model almost always returns a sub-optimal choice in some challenging scenarios, such as the boundary map and the narrow pathway map. Second, the proposed planning algorithm, namely the padding mean neural dynamic model, is a topologically organized network with connections among neighbouring neurons and is good for spreading nerve impulses such as a waves without coupling effects. The signal transduction method within a network is based on a dynamic neural activity field, which propagates high neural activity from the target state to the whole field, excluding obstacle regions. Third, simulation studies are conducted to compare the performance of the proposed planning algorithm and other popular planning algorithms in terms of effectiveness and efficiency. As a result, the proposed method can drive a robot to find more reasonable paths in both static maps and unknown dynamic scenarios with moving obstacles and a moving target. Finally, the novel excitatory input design of the proposed algorithm is discussed and analysed to explore the neural stimulus propagation mechanism within the network.